I think I've discovered a problem with Repackager and would appreciate it if anyone here could try to confirm (or otherwise) what I have found. The packaging environment is a virtualised (on VMWare) 64bit Windows 7 and I am using snapshotting.
Any registry value of type REG_QWORD gets picked up by repackager irrespective of whether or not it has changed between snapshots. Furthermore, repackager reports these values as of being of type REG_SZ.
These REG_SZ keys then get included in the repackaged .msi. So, it is conceivable that a single .msi could change EVERY (or at least, many) REG_QWORD in the registry to REG_SZ and this may lead to operating system corruption. One example of where this definitely happens is with the following key
[HKEY_LOCAL_MACHINE\SOFTWARE\Microsoft\Windows\CurrentVersion\SideBySide]
"PublisherPolicyChangeTime"=hex(b):5d,38,a3,58,e0,f9,cb,01
This key is in my repackaging image as a REG_QWORD value. If I launch repackager and perform two snapshots (without changing anything), repackager reports it as a REG_SZ value. The resulting .msi will, on installation, actually modify this key to a REG_SZ and corrupt the operating system.
I initially thought that this was a problem with a particular package (notepad++ - see
http://community.flexerasoftware.com/showthread.php?t=199811 ) but now realize that it is a more general problem.
As another demonstration, I created a REG_QWORD value at HKLM\SOFTWARE\mike_test BEFORE launching repackager
Windows Registry Editor Version 5.00
[HKEY_LOCAL_MACHINE\SOFTWARE\mike_test]
"test123"=hex(b):45,23,01,00,00,00,00,00
I then launched repackager, took two snapshots and observed that this key is reported as a REG_SZ value. I did not change it in between taking the two snapshots.
Whenever you take two snapshots on a clean system, you always expect a certain amount of operating system noise. However, it turns out that much of what I thought was 'noise' was actually just a lot of unchanged REG_QWORDs that get picked up in this manner.
Best Wishes,
Mike